Response mapping for EQ-5D-3L to EQ-5D-5L: a survey experiment on the validity of mapping data
|Day:||Thu 4 Jul|
EQ-5D is a 5-item questionnaire module measuring health-related quality of life. It is used to measure health benefits within many cost-effectiveness studies providing evidence for decisions on medical technologies made by the National Institute for Health and Care Excellence (NICE) in England and similar policy bodies in other countries. Recently redesigned in a more detailed form (EQ-5D-5L), much existing cost-effectiveness evidence remains based on the older version (EQ-5D-3L). Statistical mapping from one version to another is widely used, exploiting reference datasets containing both variants. We report a randomised experiment in the UKHLS Innovation Panel designed to investigate whether use of both versions in a single interview gives a reliable picture of the relationship between the two health measures. Comparing an experimental group that receives only EQ-5D-5L with another that receives EQ-5D-3L followed by EQ-5D-5L, we find that measurement by EQ-5D-5L gives a significantly more positive picture of health when preceded by EQ-5D-3L. The use of web rather than face-to-face interviewing has a similar significant effect. This suggests that a reference dataset carrying both versions of the EQ-5D instrument (particularly if administered online) may give a systematically different measure of health relative to a conventional interview including EQ-5D-5L on its own. Consequently, statistical mapping used to convert older EQ-5D-3L evidence into newer EQ-5D-5L form may give biased results, leading to distorted cost-effectiveness decisions. We illustrate the practical significance of this bias in relation to some real-world examples where mapping has been used.